Repository navigation
CLI: add hflow stale --exit-code so CI can gate on stale episodes #30
Description
Current behavior
_command_stale (src/hflow/cli.py:268) prints stale episode URIs and always returns 0 (src/hflow/cli.py:318), regardless of how many episodes are behind the current pipeline version. That means hflow stale cannot gate a CI job or a pre-training checklist, while hflow doctor already exits 1 on findings.
Proposed change
Add an --exit-code flag (named after git diff --exit-code): when passed, exit 1 if any stale episode was printed. Keep the default at 0 so existing pipes keep working (docs/CATALOG.md documents piping the output into xargs hflow ingest).
Definition of done
hflow stale --exit-code ...exits 1 when at least one stale episode exists and 0 when none do.- Without the flag, behavior is unchanged.
- Test with the existing catalog fixtures in
tests/test_catalog_curation.py.
Validation
uv run pytest -q
uv run ruff check --fix && uv run ruff format
uv run ty check
Activity
chiruu12 commentedon Aug 20, 2026
Happy to take this one. Plan: add
--exit-code(store_true) to thestalesubparser and return 1 from_command_stalewhen it's set and at least one stale episode was printed. Default stays 0 so thexargs hflow ingestpipe in docs/CATALOG.md is unaffected. I'll cover the behind and up-to-date cases against the existing fixtures in test_catalog_curation.py and document the flag.kstonekuan commentedon Aug 20, 2026
Yours, @chiruu12, assigned. The plan:
store_trueon thestalesubparser, return 1 from_command_staleonly when the flag is set and at least one stale episode printed, default exit code unchanged so thexargspipe in docs/CATALOG.md keeps working.kstonekuan commentedon Aug 21, 2026
Heads-up @chiruu12: #30 is still yours, but someone else opened a pull request for it (#76, now closed since you have the assignment). No pressure and no deadline, I just wanted you to know there is interest, in case you would rather hand it off than hold it.
If you have hit a snag, say so here and I will help. If you would rather do something else, that is completely fine too, just let me know and I will unassign so it goes back in the pool.
chiruu12 commentedon Aug 21, 2026
Thanks @kstonekuan, and thanks for handling #76 the way you did.
No snag, I want to keep it. The reason there was no PR yet is that I wanted the change verified rather than just written. The flag itself is a few lines, but the part worth getting right is that the default path stays exactly what it was, since
docs/CATALOG.mdpipesstaleintoxargs hflow ingestand a changed exit code there breaks that quietly. So I checked the new tests fail on currentmainwithunrecognized arguments: --exit-codebefore checking they pass on the change, and I ran the whole suite onmainas well as on the branch so the comparison is against a measured base rather than an assumed one.PR is up now: #81
Local run on macOS with
uv sync --locked --all-extras:uv run pytest -q tests/test_catalog_curation.py -k cli_stale, 4 passeduv run pytest -q, 301 passed, 5 failed, 6 skipped. The same 5 fail on unmodifiedmainon this host, all ffmpeg related, and base sits at 299 passed, so the change adds two and moves nothing elseuv run ruff checkanduv run ruff format --check, cleanuv run ty check, clean